August 1st, 2011

A Set of Calculators for Estimating Readmission Risk

If you are interested in a tool that estimates the readmission risk of a patient who has been hospitalized for acute myocardial infarction, heart failure, or pneumonia, you can find one at

Readmission rates are increasingly a focus of quality-of-care efforts in the U.S., including those initiated by the Centers for Medicare and Medicaid Services, such as the work my colleagues and I do at the Yale-New Haven Hospital Center for Outcomes Research and Evaluation. To validate models that had been based on administrative codes, we developed models using medical records information. These models do a good job of predicting readmission risk, but they were not developed for the specific purpose of creating the tool you’ll find at

So here are a few things to bear in mind as you use our risk calculators to help a patient make the transition from the hospital to home:

1. The calculators provide an estimate of risk, not a pinpointed assessment of it.

2. The calculators assume that the performance of the treating hospital is average in terms of readmission rates. Hospitals that perform better or worse than average may have readmission rates that differ accordingly.

3. When we developed our models, we did not seek to limit the number of variables (as the calculators do) or to include information about in-hospital adverse events. The models were intended to calculate risk using all of the information about the patient’s condition upon presentation to the hospital — that’s because they were assessing hospital quality, and we did not want to adjust for complications. Therefore, you should consider factors that may be important but are not included in the tool. I am hopeful that future calculators may improve over time.

4. The calculators do not say how to use the estimates. That’s up to you. It may be that high-risk patients (e.g., >30% chance of readmission within 30 days) would merit additional services to support the transition to home. Whether — and how — this information will improve care are questions that remain to be answered. Our hope is that we can work together with clinicians and other health care professionals who try this tool to ultimately determine how best to use it in practice.

5. We eventually want to turn these calculators into a mobile app. Because, unfortunately, the programming is expensive, we may decide to charge a modest amount to offset the costs. On the web, though, you can use the calculators for free. All the information we used to create them is in the public domain. We will be deciding about the mobile apps in the next couple of weeks. If you have an opinion, be sure to make a comment below.

Please make use of these free calculators as we all work to reduce readmission rates. And tell me: Is the tool helpful? How can it best be used to benefit patients?

4 Responses to “A Set of Calculators for Estimating Readmission Risk”

  1. Saurav Chatterjee, MD says:

    Thanks for the scoring tool Dr Krumholz-it is lucid, handy and quite elaborate….I presume that it has been tested with a cohort and the results have been published-would it be possible to post the studies that have used this tool effectively?

    Competing interests pertaining specifically to this post, comment, or both:

  2. Glad you like the calculators. A link to the articles that describe them are at the bottom of each data entry page.

  3. Dear Dr. Krumholz

    The heart failure risk calculator is user-friendly and practical as the variables are readily available early on in the admission period. A mobile app is a great idea.

    However the article that is referenced describes a risk measure that was developed for hospital profiling of readmission rates, rather than for risk stratification of individual patients. Death – a competing risk – was not accounted for in the outcome. As you mention above, because the variables were selected for hospital profiling, in-hospital complications were not included as variables in the model. Finally, the discrimination of the model was modest (which is the case for other readmission risk prediction tools as well, likely due to variations in process of care).

    As such, I wonder how reliable – in your opinion – the risk calculator is in assessing the readmission risk of individual patients at their index hospitalization. I am initiating an implementation project to improve the process of care for hospitalized heart failure patients throughout the hospitals in our regional care network; in the interest of cost-effectiveness, I would like to tailor the intervention to the readmission risk of individual patients. The participating hospitals vary in performance, setting, and volume of cases, and the goal of the project is to reduce 30-d readmission rates.

    With best wishes,

    Competing interests pertaining specifically to this post, comment, or both:

  4. Thank you Harriette. You are very astute. The measure was built for hospital profiling but it turns out to be about as good as most other measures that were designed for patient prediction. It is not what I would have done if the intent from the outset was patient prediction. Nevertheless, it does a pretty good job of prediction and I believe that it can be used. We are about to release an iPhone app that should make it easier to use.